Correct a comment in YCbCrConverter.

[nageru] / NEWS

Nageru and Futatabi 1.9.3, April 12th, 2020

  - Support (video-only) V4L2 output. The intended use case is output into
    v4l2loopback to get into videoconferencing or the likes:

     sudo apt install v4l2loopback-dkms
     sudo modprobe v4l2loopback video_nr=2 card_label='Nageru loopback' max_width=1280 max_height=720 exclusive_caps=1
     nageru --v4l-output /dev/video2
   
   Start Nageru before any readers.


Nageru and Futatabi 1.9.2, March 29th, 2020

  - Support handling white balance directly in Nageru, without themes
    manually inserting a WhiteBalanceEffect or handling set_wb().
    To use it, call scene:add_white_balance() instead of
    scene:add_effect(WhiteBalanceEffect.new()). If using this functionality,
    white balance will be properly propagated to the MJPEG feed and
    through Futatabi, so that replays get the correct white balance.
    Futatabi's UI will still be uncorrected, though.

  - Make it possible to siphon out a single MJPEG stream, for remote
    debugging, single-camera recording, single-camera streaming via
    Kaeru or probably other things. The URL for this is /feeds/N.mp4
    where N is the card index (starting from zero).

  - The theme can now access some audio settings; it can get (not set)
    number of buses and names, get/set fader volume, get/set mute,
    and get/set EQ parameters.

  - In Futatabi, it is now possible to set custom source labels, with
    the parameter --source-label NUM:LABEL (or -l NUM:LABEL).

  - When the playback speed changes in Futatabi, ease into the new speed.
    The easing period is nominally 200 ms, but it will be automatically
    shortened or lengthened (up to as much as two seconds in extreme
    cases, especially involving very slight speed change) if this
    helps getting back into a cadence of hitting the original frames.
    This can mean significant performance improvements when ramping
    from higher speeds back into 100%.

  - Updates for newer versions of CEF (tested with Chrome 80).

  - Various bugfixes and performance improvements.


Nageru and Futatabi 1.9.1, November 17th, 2019

  - Support disabling optional effects if a given other effect is _enabled_
    (typically for mutually exclusive effects).
    
  - Make it possible for the theme to override the status line, by declaring
    a function format_status_line() in the theme. Inspired by a C++ patch by
    Alex Thomazo in the Breizhcamp repository.

  - Various bugfixes.


Nageru and Futatabi 1.9.0, July 20th, 2019

  - Significant reworking of the theme engine: Chains (now called scenes)
    can now instantiate different versions behind-the-scenes instead of the
    user having to worry about input types, low/high quality, or replacing
    effects with others. Menus can have submenus and checkboxes. Finally, some
    callbacks, such as num_channels(), have been replaced with easier-to-use
    imperative functions, ie., an explicit call to Nageru.set_num_channels(N).

    See the documentation for more information, or the included themes,
    which have been ported to the new interfaces. Existing themes will continue
    to run without modification, but the old interfaces are deprecated.

    Martin Sandsmark contributed a bugfix to this work.

  - Support cross-compilation. Patch from Helmut Grohne.

  - Kaeru now has a parameter --disable-audio for transcoding streams
    with no audio.

  - Various bugfixes. In particular, work around an issue where Mesa's shader
    cache interacts with Qt's EGL support to create a confusing crash with
    “vertex shader lacks `main'”.


Nageru and Futatabi 1.8.6, April 19th, 2019

  - Filenames for the recordings are now without colons; it caused
    too much problems with various software, including most players.

  - Various bugfixes.


Nageru and Futatabi 1.8.5, March 30th, 2019

  - Experimental support for audio in Futatabi: The MJPEG export from
    Nageru now supports audio, and Futatabi will store it and play it
    back. Audio is currently only supported when playing at 100% speed
    (no pitch shift or time stretching), and there is no audio output
    to the Futatabi operator.

  - Significant optimizations to MJPEG encoding, both when in use and when
    not in use.

  - Various bugfixes.


Nageru and Futatabi 1.8.4, March 11th, 2019

  - Various bugfixes, in particular for 32-bit platforms.


Nageru and Futatabi 1.8.3, March 10th, 2019

  - Allow controlling video mixing from MIDI events. Adapted from a patch
    by Yann Dubreuil, from the BreizhCamp repository.

  - Use ALSA hardware timestamps for input; gives more stable delay.
    Patch by Yann Dubreuil, from the BreizhCamp repository.

  - For FFmpeg inputs, add an option for playing as fast as possible
    (set rate >= 10.0).

  - In Futatabi, support queueing and playing clips with no cue-out point.
    This opens up for new and even faster UI workflows.

  - Many bugfixes.


Nageru and Futatabi 1.8.2, January 19th, 2019

  - Futatabi now supports MIDI controllers like Nageru, including an editor
    and a sample mapping for the Behringer CMD PL-1.

  - Futatabi now supports changing master speed during play, both via a
    MIDI controller and the GUI.

  - Various bugfixes.


Nageru and Futatabi 1.8.1, December 30th, 2018

  - Futatabi can now communicate its queue status through a subtitle track,
    and Nageru can consume it. This allows Nageru themes to get precise
    information programmatically, e.g. to show status or automatically
    switch away when the queue is about to end.

  - Futatabi can now reuse the computed flow across successive frames when
    interpolating between the same frame pair. This significantly reduces
    the GPU load when doing super-slow motion (slower than 0.5x).

  - Various smaller fixes.


Nageru and Futatabi 1.8.0, December 20th, 2018

  - Initial release of Futatabi, a multicamera slow motion video server
    designed to be used with Nageru. Futatabi is currently in alpha stage
    and largely undocumented.

  - Add support for multi-camera export from Nageru. A multi-camera stream
    contains all frames from all camera inputs (unless overridden by
    --mjpeg-export-cards), unprocessed except for MJPEG encoding.
    MJPEG encoding is done in hardware (via VA-API) on Skylake or newer,
    or using libjpeg otherwise. The intended user of this stream is Futatabi.


Nageru 1.7.5, November 11th, 2018

  - Fix a bug where --record-x264-video would not work when VA-API was
    not present, making the option rather useless (broken in 1.7.2).
    Bug reported by Peter De Schrijver.

  - The build system has been switched to Meson; see the README for new
    build instructions.

  - Various smaller fixes.


Nageru 1.7.4, August 31st, 2018

  - Rework the x264 speedcontrol presets, again. (They earlier assumed
    we could control B-frame settings on the fly, which we cannot with
    threaded lookahead.) Also support x264 >= 153, which can support
    multiple bit depths in the same library.

  - Default to SDI inputs instead of HDMI.

  - Add a mode to run in full screen (--fullscreen). Adapted from a patch
    by Yoann Dubreuil.

  - Add support for lift/gamma/gain in the theme. Patch by Alexandre Thomazo.


Nageru 1.7.3, May 22nd, 2018

  - When using multichannel audio, add a control for adjusting the
    stereo width (from normal stereo to mono, all the way to
    inverted stereo).

  - Removed --http-coarse-timebase (it is now always on).

  - Various bugfixes.


Nageru 1.7.2, April 28th, 2018

  - Several improvements to video (FFmpeg) inputs: You can now use
    them as audio sources, you can right-click on video channels
    to change URL/filename on-the-fly, themes can ask for forced
    disconnection (useful for network sources that are hanging),
    and various other improvements. Be aware that the audio support
    may still be somewhat rough, as A/V sync of arbitrary video
    playout is a hard problem.

  - The included themes have been fixed to properly make the returned
    chain preparation functions independent of global state (e.g. if
    the white balance for a channel was changed before the frame was
    actually rendered). If you are using a custom theme, you may want
    to apply similar fixes to it.

  - In Metacube stream output, mark each keyframe with a pts metadata
    block. This allows Cubemap 1.4.0 or newer to serve fMP4 fragments
    for HLS from Nageru's output, without any further remuxing or
    transcoding.

  - If needed, Nageru will now automatically try to autodetect a
    usable --va-display parameter by probing all DRM nodes for H.264
    encoders. This removes the need to set --va-display in almost all
    cases, and also removes the dependency on libpci.

  - For GPUs that support querying available memory (in practice only
    NVIDIA GPUs at the current time), expose the amount of used/total
    GPU memory both on standard output and in the Prometheus metrics
    (as well as included Grafana dashboard).

  - The Grafana dashboard now supports heatmaps for the chosen x264
    speedcontrol preset (requires Grafana 5.1 or newer). (There used to
    be a heatmap earlier, but it was all broken.)

  - Various bugfixes.


Nageru 1.7.1, March 26th, 2018

  - Various bugfixes, mostly related to HTML and video inputs.


Nageru 1.7.0, March 8th, 2018

  - Support for HTML5 graphics directly in Nageru, through CEF
    (Chromium Embedded Framework). This performs better and is more
    flexible than integrating with CasparCG over a socket. Note that
    CEF is an optional component; see the documentation for more
    information.

  - Add an HTTP endpoint for enumerating channels and one for getting
    only their colors. Intended for remote tally applications;
    set the documentation.

  - Add a video grid display that removes the audio controls and shows
    the video channels only, potentially in multiple rows if that makes
    for a larger viewing area.

  - Themes can now present simple menus in the Nageru UI. See the
    documentation for more information.

  - Various bugfixes.


Nageru 1.6.4, January 25th, 2018

  - Fix compilation with the upcoming FFmpeg 3.5.

  - Switch to LuaJIT for the theme engine, which is faster.

  - Various bugfixes and smaller optimizations.


Nageru 1.6.3, November 8th, 2017

  - Add quick-cut keys (Q, W, E, etc.) below the preview keys.
    Since it's easy to hit these by accident and put up a signal
    you didn't want, they are disabled by default (they can be
    enabled in the video menu, or with the command line flag
    --quick-cut-keys).

  - Rework the x264 speedcontrol presets to better match newer
    x264 versions.

  - Add an option for changing the HTTP port (--http-port).

  - Various smaller bug and integration fixes.


Nageru 1.6.2, July 16th, 2017

  - Various smaller Kaeru fixes, mostly around metrics. Also,
    you can now adjust the x264 bitrate in Kaeru (in 100 kbit/sec
    increments) by sending SIGUSR1 (higher) or SIGUSR2 (lower).


Nageru 1.6.1, July 9th, 2017

  - Add native export of Prometheus metrics.

  - Rework the frame queue drop algorithm. The new one should handle tricky
    situations much better, especially when a card is drifting very slowly
    against the master timer.

  - Add Kaeru, an experimental transcoding tool based on Nageru code.
    Kaeru can run headless on a server without a GPU to transcode a
    Nageru stream into a lower-bitrate one, replacing VLC.

  - Work around a bug in some versions of NVIDIA's OpenGL drivers that would
    crash Nageru after about three hours (fix in cooperation with Movit).

  - Fix a crash with i965-va-driver 1.8.x.

  - Reduce mutex contention in certain critical places, causing lower tail
    latency in the mixer.


Nageru 1.6.0, May 29th, 2017

  - Add support for having videos (from file or from URL) as a separate
    input channels, albeit with some limitations. Apart from the obvious use of
    looping pause clips or similar, this can be used to integrate with CasparCG;
    see the manual for more details.

  - Add a frame analyzer (accessible from the Video menu) containing an
    RGB histogram and a color dropped tool. This is useful in calibrating
    video chains by playing back a known signal. Note that this adds a
    dependency on QCustomPlot.

  - Allow overriding Y'CbCr input interpretation, for inputs that don't
    use the correct settings. Also, Rec. 601 is now used by default instead
    of Rec. 709 for SD resolutions.

  - Support other sample rates than 48000 Hz from bmusb.


Nageru 1.5.0, April 5th, 2017

  - Support for low-latency HDMI/SDI output in addition to (or instead of) the
    stream. This currently only works with DeckLink cards, not bmusb. See the
    manual for more information.

  - Support changing the resolution from the command line, instead of locking
    everything to 1280x720.

  - The A/V sync code has been rewritten to be more in line with Fons
    Adriaensen's original paper. It handles several cases much better,
    in particular when trying to match 59.94 and 60 Hz sources to each other.
    However, it might occasionally need a few extra seconds on startup to
    lock properly if startup is slow.

  - Add support for using x264 for the disk recording. This makes it possible,
    among other things, to run Nageru on a machine entirely without VA-API
    support.

  - Support for 10-bit Y'CbCr, both on input and output. (Output requires
    x264 disk recording, as Quick Sync Video does not support 10-bit H.264.)
    This requires compute shader support, and is in general a little bit
    slower on input and output, due to the extra amount of data being shuffled
    around. Intermediate precision is 16-bit floating-point or better,
    as before.

  - Enable input mode autodetection for DeckLink cards that support it.
    (bmusb mode has always been autodetected.)

  - Add functionality to add a time code to the stream; useful for debugging
    latency.

  - The live display is now both more performant and of higher image quality.

  - Fix a long-standing issue where the preview displays would be too bright
    when using an NVIDIA GPU. (This did not affect the finished stream.)

  - Many other bugfixes and small improvements.


Nageru 1.4.2, November 24th, 2016

  - Fix a thread race that would sometimes cause x264 streaming to go awry.


Nageru 1.4.1, November 6th, 2016

  - Various bugfixes.


Nageru 1.4.0, October 26th, 2016

  - Support for multichannel (or more accurately, multi-bus) audio,
    choosable from the UI or using the --multichannel command-line
    flag. In multichannel mode, you can take in inputs from multiple
    different sources (or different channels on the same source, for
    multichannel sound cards), apply effects to them separately and then
    mix them together. This includes both audio from the video cards
    as well as ALSA inputs, including hotplug. Ola Gundelsby contributed
    invaluable feedback on this feature throughout the entire
    development cycle.

  - Support for having MIDI controllers control various aspects of the
    audio UI, with relatively flexible mapping. Note that different
    MIDI controllers can vary significantly in what protocol they speak,
    so Nageru will not necessarily work with all. (The primary testing
    controller has been the Akai MIDImix, and a pre-made mapping for
    that is included. The Korg nanoKONTROL2 has also been tested and
    works, but it requires some Korg-specific SysEx commands to make
    the buttons and lights work.)

  - Add a disk space indicator to the main window.

  - Various bugfixes. In particular, an issue where the audio would pitch
    up sharply after a series of many dropped frames has been fixed.


Nageru 1.3.4, August 2nd, 2016

  - Various bugfixes.


Nageru 1.3.3, July 27th, 2016

  - Various changes to make distribution packaging easier; in particular,
    theme data can be picked up from /usr/local/share/nageru.

  - Fix various FFmpeg deprecation warnings, now that we need FFmpeg
    3.1 for other reasons anyway.


Nageru 1.3.2, July 23rd, 2016

  - Allow limited hotplugging (unplugging and replugging) of USB cards.
    You can use the new command-line option --num-fake-cards (-C) to add
    fake cards that show only a single color and that will be replaced
    by real cards as you plug them in; you can also unplug cards and have
    them be replaced by fake cards. Fake cards can also be used for testing
    Nageru without actually having any video cards available.

  - Add Metacube timestamping of every keyframe, for easier detection of
    streams not keeping up. Works with the new timestamp feature of
    Cubemap 1.3.1. Will be ignored (save for some logging) in older
    Cubemap versions.

  - The included default theme has been reworked and cleaned up to be
    more understandable and extensible.

  - Add more command-line options for initial audio setup.


Nageru 1.3.1, July 1st, 2016

 - Various display bugfixes.


Nageru 1.3.0, June 26th, 2016

 - It is now possible, given enough CPU power (e.g., a quad-core Haswell or
   faster desktop CPU), to output a stream that is suitable for streaming
   directly to end users without further transcoding. In particular, this
   includes support for encoding the network stream with x264 (the stream
   saved to disk is still done using Quick Sync), for Metacube framing (for
   streaming to the Cubemap reflector), and for choosing the network stream
   mux. For more information, see the README.

 - Add a flag (--disable-alsa-output) to disable ALSA monitoring output.

 - Do texture uploads from the main thread instead of from separate threads;
   may or may not improve stability with NVIDIA's proprietary drivers.

 - When beginning a new video segment, the shutdown of the old encoder
   is now done in a background thread, in order to not disturb the external
   stream. The audio still goes into a somewhat random stream, though.

 - You can now override the default stream-to-card mapping with --map-signal=
   on the command line.

 - Nageru now tries to lock itself into RAM if it has the permissions to do
   so, for better realtime behavior. (Writing the stream to disk tends to
   fill the buffer cache, eventually paging less-used parts of Nageru out.)

 - Various fixes for deadlocks, memory leaks, and many other errors.


Nageru 1.2.1, April 15th, 2016

 - Images are now updated from disk about every second, so that it is possible
   to update e.g. overlays during streaming, although somewhat slowly.

 - Fix support for PNG images.

 - You can now send SIGHUP to start a new cut instead of using the menu.

 - Added a --help option.

 - Various tweaks to OpenGL fence handling.


Nageru 1.2.0, April 6th, 2016

 - Support for Blackmagic's PCI and Thunderbolt cards, using the official
   (closed-source) Blackmagic drivers. (You do not need the SDK installed, though.)
   You can use PCI and USB cards pretty much interchangeably.

 - Much more stable handling of frame queues on non-master cards. In particular,
   you can have a master card on 50 Hz and another card on 60 Hz without getting
   lots of warning messages and a 10+ frame latency on the second card.

 - Many new options in the right click menu on cards: Adjustable video inputs,
   adjustable audio inputs, adjustable resolutions, ability to select card for
   master clock.

 - Add support for starting with almost all audio processing turned off
   (--flat-audio).

 - The UI now marks inputs with red or green to mark them as participating in
   the live or preview signal, respectively. Red takes priority. (Actually,
   it merely asks the theme for a color for each input; the theme contains
   the logic.)

 - Add support for uncompressed video instead of H.264 on the HTTP server,
   while still storing H.264 to files (--http-uncompressed-video). Note that
   depending on your client, this might not actually be more CPU efficient
   even on localhost, so be sure to check.

 - Add a simpler, less featureful theme (simple.lua) that should be easier to
   understand for beginners. Themes are now also choosable with -t on the command
   line.

 - Too many bugfixes and small tweaks to list. In particular, many memory leaks
   in the streaming part have been identified and fixed.


Nageru 1.1.0, February 24th, 2016

 - Support doing the H.264 encoding on a different graphics device from the one
   doing the mixing. In particular, this makes it possible to use Nageru on an
   NVIDIA GPU while still encoding H.264 video using Intel Quick Sync (NVENC
   is not supported yet) -- it is less efficient since the data needs to be read
   back via the CPU, but the NVIDIA cards and drivers are so much faster that it
   doesn't really matter. Tested on a GTX 950 with the proprietary drivers.

 - In the included example theme, fix fading to/from deinterlaced sources.

 - Various smaller compilation, distribution and documentation fixes.


Nageru 1.0.0, January 30th, 2016

 - Initial release.